Identification of the Catalytic Ubiquinone-binding Site of Vibrio cholerae Sodium-dependent NADH Dehydrogenase

The sodium-dependent NADH dehydrogenase (Na+-NQR) is a key component of the respiratory chain of diverse prokaryotic species, including pathogenic bacteria. Na+-NQR uses the energy released by electron transfer between NADH and ubiquinone (UQ) to pump sodium, producing a gradient that sustains many essential homeostatic processes as well as virulence factor secretion and the elimination of drugs. The location of the UQ binding site has been controversial, with two main hypotheses that suggest that this site could be located in the cytosolic subunit A or in the membrane-bound subunit B. In this work, we performed alanine scanning mutagenesis of aromatic residues located in transmembrane helices II, IV, and V of subunit B, near glycine residues 140 and 141. These two critical glycine residues form part of the structures that regulate the site's accessibility. Our results indicate that the elimination of phenylalanine residue 211 or 213 abolishes the UQ-dependent activity, produces a leak of electrons to oxygen, and completely blocks the binding of UQ and the inhibitor HQNO. Molecular docking calculations predict that UQ interacts with phenylalanine 211 and pinpoints the location of the binding site in the interface of subunits B and D. The mutagenesis and structural analysis allow us to propose a novel UQ-binding motif, which is completely different compared with the sites of other respiratory photosynthetic complexes. These results are essential to understanding the electron transfer pathways and mechanism of Na+-NQR catalysis.

In contrast with the proton gradient used by eukaryotic mitochondria, diverse prokaryotic species employ a sodium gradient to sustain crucial homeostatic processes, such as nutrient transport, ATP synthesis, flagellum rotation, and pH regulation (1,2). Many pathogenic bacteria also use the sodium gradient to drive processes involved in the secretion of virulence factors and the efflux of drugs, enabling antibiotic resistance (1)(2)(3)(4). The sodium-dependent NADH dehydrogenase (Na ϩ -NQR) 2 is a respiratory enzyme that couples the electron transfer from NADH to ubiquinone (UQ) with sodium pumping (5)(6)(7)(8) and seems to have a major role in the production of the sodium gradient in many types of bacteria (1)(2)(3).
The Na ϩ -NQR complex consists of six subunits (A-F) (Fig.  1A) and five confirmed redox cofactors: FAD, 2Fe-2S center, two covalently bound FMN cofactors, and, remarkably, riboflavin (5)(6)(7)(8). Na ϩ -NQR is the only reported enzyme that is able to use riboflavin directly as a redox carrier (9,10) as all other flavoproteins use FMN or FAD (11). Subunit A is a cytosolic protein with no cofactors but, interestingly, contains motifs reminiscent of NAD and ferredoxin binding domains, indicating its possible origins in the homologous subunit C of the RNF complex (a membrane-bound ferredoxin oxidoreductase) (2). Subunit B contains 10 transmembrane segments and one of the two covalently bound FMN cofactors (12)(13)(14). This subunit also plays an important role in sodium transport and carries one of the at least two sodium binding sites and an ion channel (15)(16)(17)(18). Subunit C has a large periplasmic domain that contains the second covalently bound FMN cofactor and a single transmembrane segment (12)(13)(14). Subunits D and E are homologous proteins with six transmembrane helices, with an antiparallel orientation and contain the second sodium binding site (12,16,18,19). The crystal structure suggests that these two subunits also have a spectroscopically silent iron center, bound through four conserved cysteine residues (7,12). However, the presence of this center requires further experimental confirmation. Subunit F has a large cytosolic domain that contains the binding sites for NADH, FAD, and the 2Fe-2S center and a transmembrane segment (20).
The internal electron transfer sequence, the redox and spectral properties of all confirmed cofactors, and the redox steps involved in sodium transport have been extensively characterized (21)(22)(23)(24)(25)(26). The data indicate that the electrons are shuttled from NADH to UQ, through a linear "downhill" pathway (21,22). FAD is the first cofactor of the sequence, accepting two electrons from NADH (21). The electrons are then transferred, one by one, to the 2Fe-2S center, to the two FMN molecules, and to riboflavin, which delivers the redox equivalents to UQ (10,21). Although the UQ binding site is involved in the last step of the electron transfer pathway, it has remained largely uncharacterized.
Although the recently published crystallographic data (12) have illuminated different structural characteristics and have aided in understanding the operation of this complex, specific aspects of Na ϩ -NQR function are unknown. One of the unanticipated findings in the structure of Na ϩ -NQR is the absence of UQ or a cavity that could accommodate this molecule. The purified Na ϩ -NQR complex contains a tightly bound UQ molecule (27,28) that does not seem to have a specific physiologic role, and a separate UQ binding site that actually participates in the catalytic cycle (29,30). The location of the catalytic UQ binding site has been elusive, with two main hypotheses that locate the site in either subunit A or B. Studies by Casutt et al. (31) have shown that subunit A is labeled with photoreactive biotinylated UQ analogs and that this subunit could bind up to two UQ molecules (31,32). Our group has followed the early proposal by Unemoto's group (33,34), which suggests that the UQ binding site could be located in subunit B. Hayashi et al. (6,14) described spontaneous Vibrio harveyi mutants that were resistant to the inhibitors N-oxo-2-heptyl-4-hydroxyquinoline (HQNO) and korormicin, which are ubiquinone analogs. The inhibitor-resistant strains contained a single mutation in conserved glycine residue 141 (Vibrio cholerae numbering) of subunit B, suggesting that this residue could form part of the UQ binding site. Our group further studied this site and characterized mutants of Gly-141 and the contiguous conserved glycine residue 140. Our studies confirmed that the mutations in position 141 decreased the susceptibility toward the inhibitors. Moreover, we were able to establish that mutations of Gly-140 abolished the UQ-dependent activity and increased the K m for UQ by several orders of magnitude (29). Kinetic analysis showed that the specific step that is altered in these mutants is the reduction of UQ (29), the final step of the catalytic cycle. IR spectroscopy studies showed that the conformational changes involved in the redox-induced binding of UQ are blocked in these mutants (30). However, UQ affinity was not altered, indicating that these residues do not participate directly in the binding site but probably regulate the redox-induced accessibility (29,30). Although the evidence clearly supports the presence of two UQ binding sites, understanding of their specific roles, locations, and possible interactions is required to elucidate the mechanism of Na ϩ -NQR.
In this work, we performed alanine scanning mutagenesis of conserved aromatic residues in the vicinity of the two key glycine residues 140 and 141. Aromatic residues could participate in the UQ binding site, formingstacking interactions with the benzoquinone head or stabilizing the semiquinone radicals formed during UQ reduction. The data show that the mutants of residues in transmembrane helices II, IV, and IV decrease the UQ-dependent activity and increase the leak of electrons to oxygen, consistent with an important role of these residues in the UQ binding site. Mutants F211A and F213A, which showed the more pronounced effects, were further characterized. In contrast with the behavior of the mutants of residues Gly-140 and Gly-141, the phenylalanine mutants completely blocked the ability of the enzyme to bind ubiquinone, showing unsaturable behavior for this substrate, and decreased the binding of the inhibitor HQNO by several orders of magnitude, strongly indicating that these residues do indeed participate in the catalytic binding site. Molecular docking calculations independently pinpoint a plausible location for this site in the interface of subunits B and D, which is now allowing us to map the site and understand the role of each residue in electron transfer and substrate binding. The data presented in this report help to clarify the structure-function relationships of Na ϩ -NQR, provide structural information about a novel UQ-binding motif, and are critical to understanding the catalytic mechanism of this essential respiratory complex.

Results
Alanine scanning mutagenesis was performed on conserved aromatic residues in transmembrane helix (TH) II, TH IV, and TH V of subunit B (Fig. 1, B-D). These residues are in the vicinity of glycine residues 140 and 141, found in TH II. The alanine mutants of these two glycine residues show an increase in the K m for UQ-1 and block the conformational changes involved in its binding but, remarkably, do not change the affinity of the enzyme for the substrate, indicating that they regulate the formation or opening of the UQ binding site but do not form part of the actual site (29,30). Aromatic residues were selected as possible parts of the UQ site due to the potentialstacking interactions that could help bind this substrate, which makes them common components in the binding sites of other respiratory and photosynthetic complexes (35)(36)(37)(38)(39). Moreover, these residues could stabilize radical species that are formed during the reduction of UQ. All confirmed redox cofactors of Na ϩ -NQR form one-electron reduced states during the catalytic cycle (21,40), and several of them can be found as stable flavin semiquinone radicals in the fully reduced and oxidized forms of the enzyme (13,(21)(22)(23). The full reduction of UQ occurs in two separate one-electron transfers and should produce semiquinone species, because riboflavin, the last cofactor in the transfer pathway, participates exclusively as a one-electron carrier (10,21). Thus, it is possible that aromatic residues could be involved both in the binding of UQ and in the stabilization of short lived semiquinone radicals.
Alanine Scanning Mutagenesis-The aromatic residues were substituted by alanine, which mainly modifies the size but does not alter the non-polar nature of the residue. The residues that were mutated include Trp-177, Phe-185, Phe-211, Phe-213, Phe-214, and Tyr-216. All of these positions are conserved among very diverse bacterial lineages, such as Bacteroidetes, Chlorobi, Chlamydiae, and proteobacteria, except for Tyr-216, in which phenylalanine can be also found, especially in Chlamydiae (Fig. 1E). To characterize the participation of these residues in the catalytic mechanism of the Na ϩ -NQR complex, the enzymatic activities were measured in the alanine mutants.
Na ϩ -NQR has three modules that can act semi-independently. Although the physiologic activity of Na ϩ -NQR is the electron transfer from NADH to UQ (UQ reductase; UQRED). This enzyme is also able to oxidize NADH using oxygen as an electron acceptor (NADH oxidase; NADH OX ) when ubiquinone is absent or in conditions in which the electron transfer chain is blocked (18). These two activities are coupled to the NADH dehydrogenase (NADH DH ) module that feeds electrons to the UQ or oxygen pathways. The NADH DH , NADH OX , and UQ RED activities were measured in the aromatic residue mutants under nearly saturating concentrations of the three substrates (50 M UQ-1, 250 M NADH, and 50 mM NaCl) and were compared with the wild-type complex. Table 1 shows that the mutants have a substantially inhibited UQ RED activity, whereas the NADH DH activity remains constant. As can be expected, the NADH OX activity was higher in the mutants that show a decreased UQ reduction rate. The data suggest that the mutations do not produce a general destabilization of the Na ϩ -NQR complex but rather that they block specific points in the electron transfer to UQ. The largest effects were produced by F211A and F213A, inhibiting the activity by 95 and 70%, respectively. These mutants were selected for further characterization.
Kinetic Characterization of F211A and F213A-To understand the role of these residues in the activity of the complex, a detailed kinetic characterization of the mutants F211A and F213A was performed. The activities were measured at different concentrations of the three substrates NADH, NaCl, and UQ-1, under nearly saturating concentrations of the other two, which allowed us to determine the K m(app) for each substrate. Because these mutants showed a small UQ RED activity, the K m(app) for sodium and NADH were measured following the NADH DH reaction, which is absolutely dependent on NADH and can be stimulated 50% by sodium (15).
The kinetic characterization indicates that F211A and F213A mutants specifically weaken the interaction with UQ-1 and have no effect on the affinity of sodium and NADH, with a similar behavior compared with the G140A and G141A mutants. The mutants had no effects on the apparent affinity for sodium or NADH under steady state conditions, because the K m(app) for these substrates was similar to that of wild-type Na ϩ -NQR (Table 2). On the other hand, F211A and F213A show unsaturable behavior when UQ-1 is used as a substrate (Fig. 2, A and B), which is similar to G140A, as reported previ-

TABLE 1 Activities of aromatic residue mutants
The activities were measured, as indicated under "Experimental Procedures," at a fixed concentration of the three substrates (250 M NADH, 50 mM NaCl, and 50 M UQ). For this work, the previously characterized mutants G140A and G141A (28,29) were constructed using as template the entire nqr operon, cloned in pBad-HisB plasmid (24). *, p Ͻ 0.05 (n Ͼ 5). ously (29). The linear response obtained indicates that the affinity of the mutants for UQ-1 decreases by several orders of magnitude (K m Ͼ100 M) compared with the wild-type enzyme, explaining the low turnover rate obtained with 50 M UQ-1 (Table 1). It should be pointed out that UQ-1 at a concentration Ͼ60 M produces substrate inhibition in these conditions (not shown), so higher concentrations were not tested in these experiments. In addition, the apparent K i for the inhibitor HQNO was measured in these mutants (Fig. 2C). F211A and F213A were less sensitive to the inhibitor HQNO than wild type, with a 25-40-fold increase in the K i ( Table 2 and Fig. 2C), which is even higher than the previously reported K i values for G140A and G141A (29,30). These results suggest that the affinity of F211A and F213A for UQ-1 and its structural analogs is several orders of magnitude smaller compared with the wildtype enzyme. Flavin Content and Stable Redox States of Na ϩ -NQR-Na ϩ -NQR contains four flavin cofactors, as demonstrated by Barquera et al. (9). All four of these flavins are found as semiquinone radical species in different stages of the catalytic cycle (21,40), and two of them produce stable flavin semiquinone radical sig-nals (10,13). In the fully oxidized state, riboflavin is found as a neutral radical (10,13), and in the fully reduced state, the FMN cofactor located in subunit B contains an anionic radical (13).
It is possible that these mutants could decrease the enzyme's activity by interfering or modifying the environment of the flavin cofactors, in particular riboflavin, which is the electron donor to UQ, rather than by affecting the UQ binding site itself. An analysis of the flavin content and the signal of the stable flavin semiquinone radicals was performed to corroborate the flavin content and reduction pattern of the mutants. Fig. 3 shows the NADH-reduced-minus-oxidized spectra of the mutants (Fig. 3, B-E), which in all cases are nearly identical to the wild-type Na ϩ -NQR spectrum, with the characteristic minima at 390, 460, and 575 nm (21,22,24). To corroborate the presence of all redox cofactors, the spectra obtained were fitted to a five-component system, comprising the full reduction of the Na ϩ -NQR complex, which contains the two-electron reductions of FAD and FMN in subunit C and one-electron reductions of the 2Fe-2S center, the neutral riboflavin cofactor, and FMN in subunit B (22)(23)(24). Table 3 shows that the reduction of all mutants follows the same pattern as the wild-type enzyme, with the expected redox transitions at the expected molar ratio, ϳ1. The reduction of the 2Fe-2S center consistently showed a smaller molar ratio of reduction (0.4 -0.7), which could be due to the relatively small signal of this cofactor, which is easily masked by the flavin signals (Fig. 3A).
In addition, the ability of riboflavin to accept electrons from UQ-1H 2 was evaluated in the mutants and wild-type Na ϩ -

NQR.
We have previously demonstrated that the reverse reduction of Na ϩ -NQR, with a molar excess of UQ-1H 2 , is incomplete, and only the riboflavin cofactor is able to accept electrons, mostly due to its relatively high midpoint potential (10,22). The UQ-1H 2 -reduced-minus-oxidized spectra (Fig. 3, G-I) show the characteristic peaks of riboflavin reduction, with minima at 500, 525, and 575 nm (10,21,22,24). Although major effects can be expected in the kinetics of reduction of riboflavin (millisecond time range), as shown for G140A (29), it is clear that mutants F211A and F213A do not interfere with the reduction of the riboflavin cofactor under equilibrium conditions (1-h incubation). Moreover, the flavin content in the mutants remained unchanged compared with the wild-type enzyme. Thus, the altered kinetic properties of the mutants are not related to a difference in cofactor content or redox properties.
Ubiquinone Binding Assays-To determine whether Phe-211 and Phe-213 participate in the UQ site, binding assays were performed in the mutants using the equilibrium dialysis methods reported previously (30). The binding experiments were carried out with samples obtained after two purification steps with DDM or after a wash with the zwitterionic detergent LDAO. The native purified Na ϩ -NQR complex obtained with DDM contains a tightly bound UQ-8, which is non-catalytic and can be eliminated with LDAO (27). The mutant G140A is able to bind a single UQ-1 or HQNO molecule (Fig. 4, B and F) with a similar affinity compared with wild-type Na ϩ -NQR (Fig.  4, A and E) ( Table 4). As reported previously (30), this relatively low affinity (micromolar) binding probably corresponds to the interactions with the catalytic UQ site. The wash with LDAO opens the non-catalytic high affinity site and increases UQ and HQNO binding to 2 molecules/mol of enzyme (30) in both wild-type Na ϩ -NQR and G140A mutant (Fig. 4, A, B, E, and F).
On the other hand, the mutants F221A and F213A completely abolished the ability of the enzyme to interact with the substrate (Fig. 4, C and D) and inhibitor (Fig. 4, G and H), decreasing the affinity of the catalytic binding site by several orders of magnitude with little to no effect on the non-catalytic high affinity site. It was previously shown that the mutants G140A and G141V, which decrease the UQ RED activity and increase the K m , showing unsaturable behavior for UQ under steady-state conditions, do not alter the affinity of the enzyme for this substrate. Due to this behavior, we proposed that these residues could control the opening of the UQ site but do not participate directly in it (29,30). The evidence found in this work with F211A and F213A strongly indicates that these residues are essential ligands forming part of the actual catalytic UQ binding site.
Molecular Docking-As an independent approach to locate the UQ binding site, we performed molecular docking calcula-tions to predict the binding pose and affinity of UQ-1 to the entire surface of subunits B, D, and E. Molecular docking predicts four low energy poses for UQ-1 at the interface of subunits B and D (Fig. 5A), with interaction energies of Ϫ98.6, Ϫ89.4, Ϫ88.1, and Ϫ85.8 kJ/mol. These poses are in the same region of the protein but appear rotated in the pocket (Fig. 5, C-F). The next lowest energy pose, located along the bottom of the interface, was scored with an interaction energy that is 8 kJ/mol weaker. Thus, molecular docking predicts that the BD interface has the general geometric characteristics and physiochemical interactions to allow UQ binding and that this substrate will bind more tightly with the interface of subunits B and D than any other part of subunits B, D, and E.
The predicted location of the UQ binding site in the interface of subunits B and D is consistent with our mutagenesis studies. The first and fourth lowest energy poses form hydrophobic  contacts with Phe-185 and Phe-211 on subunit B, which we identified as important for UQ reduction (Table 1 and Fig. 5, C and F). Compared with Phe-211, the crystal structure shows that Phe-213 is located in the opposite side of TH IV and does not face the B/D interface. Because it is unlikely for both residues to directly contact with the UQ head, it is possible that they establish interactions with different parts of the molecule, including the isoprenoid chain. Alternatively, the F213A mutant may leave a cavity that causes the helix to rearrange, disrupting the binding site. Moreover, it is possible that the pocket that we have identified forms part of the superficial location where UQ is initially bound, and in later stages of the catalytic cycle, the binding site opens fully and brings the UQ head closer to the redox cofactors.

Discussion
Location of Na ϩ -NQR Catalytic UQ Binding Site, a Novel UQ Binding Motif-The catalytic mechanism of Na ϩ -NQR has been extensively studied, and the locations of most ligand and cofactor binding motifs have been identified, through site-directed mutagenesis, functional studies (5,8,16,17,20), and X-ray crystallography (12). Moreover, the electron transfer sequence and the segments of the catalytic cycle involved in sodium transfer have been thoroughly characterized (16,18,(21)(22)(23)(24)26). The first step of the catalytic cycle is the two-electron reduction of FAD, using NADH as electron donor (20). The binding of a second NADH molecule triggers the splitting of the electron pair, and one electron is transferred to the Fe-2S center, whereas the other remains in FAD, producing a transient neutral flavin semiquinone radical (40). The subsequent redox steps are also one-electron transfer events that include the reduction of the two FMN molecules in subunits C and B (in this order), which deliver the electrons to riboflavin, the final electron carrier in the sequence (10,21). Riboflavin in turn transfers the electrons to UQ (10,21). Although the UQ binding site is involved in the critical last step of the electron transfer pathway, it has remained largely uncharacterized, with two main hypotheses regarding its location, in subunits A and B. Casutt et al. (31) have shown that subunit A is specifically labeled by photoreactive biotinylated UQ. Moreover, SPR studies demonstrated that the purified A subunit is able to bind UQ, showing a saturating kinetic behavior that suggests the binding of up to two UQ molecules (31). The reported submicromolar affinity is consistent with subunit A as the location of the tightly bound UQ, which does not seem to have a role in the catalytic mechanism, because it is found in substoichiometric amounts and does not participate in redox transfer (21)(22)(23)29), and its elimination with LDAO does not inactivate Na ϩ -NQR (27). On the other hand, the original studies by Hayashi et al. (6,14), with spontaneous V. harveyi mutants, showed that the mutation of glycine residue 141 in subunit B conferred resistance to the inhibitors korormicin and HQNO (33,34). These two inhibitors are structural analogous of UQ, suggesting that Gly-141 could form part of the UQ binding site. However, our recent results showed that HQNO is not a purely competitive inhibitor versus UQ. Instead, HQNO is a mixed type inhibitor that does not interfere with UQ binding (25), which greatly complicates the interpretation of these results. Our previous studies corroborated the data of Hayashi et al. (6,14) regarding the role of glycine residue 141 and also demonstrated that the mutation of the contiguous conserved glycine 140 almost completely abolished the UQ-dependent activity (29) and blocked the con-  formational changes induced by UQ binding (30). Nonetheless, G140A and G141A do not interfere with the binding of UQ, suggesting that these residues do not participate directly in the binding site but might regulate its aperture (30).
In this work, we used alanine scanning mutagenesis of conserved aromatic residues in the vicinity of Gly-141 and Gly-140 to locate and map the catalytic UQ binding site. The mutants that produced the largest effects, entirely blocking the UQ-dependent activity, are specifically located in the interface of subunits B and D. In particular, the mutants of residues Phe-211 and Phe-213 display non-saturating kinetics versus UQ and a 25-40-fold decrease in the sensitivity toward HQNO. In contrast with G140A and G141A, the mutants F211A and F213A completely abolished the binding of UQ-1, consistent with an almost complete lack of interaction of the substrate with the binding site. Thus, the results indicate that the elusive catalytic UQ site is located in the interface of subunits B and D and that these two phenylalanine residues play critical roles in this site. They may either form non-covalent interactions that allow the binding of UQ or stabilize the region surrounding the binding site. It should be pointed out that the kinetic parameters of the mutants F211A and F213A, with NADH and sodium as substrates, remained unaltered, indicating that the blockage in electron transfer in upstream cofactors is not involved in the decrease of activity. Moreover, the reduction pattern of the mutant enzymes showed a similar behavior compared with wild-type Na ϩ -NQR, including the reduction of the neutral riboflavin radical, which is the electron donor to UQ. Taken together, the data indicate that the mutations do not produce a general destabilization of the complex but rather specific impairment of the UQ site. Indeed, previous IR spectroscopy studies have shown that the mutation of Gly-140 does not perturb the general structure of Na ϩ -NQR (30). Although the evidence indicates that the mutations specifically alter the UQ site, it is possible that the large changes in residue size induced by the mutations (from Phe to Ala) could locally rearrange the site, preventing the formation of interactions with UQ. Further studies that incorporate bulky residues in these positions are necessary to clarify the participation of Phe-211 and Phe-213 as ligands of this substrate.
As a complementary approach to identifying the UQ site, based on a completely independent methodology, molecular docking predicted that UQ binds best to the interface between subunits B and D. This method not only identified the same region of the protein as being important to UQ binding but also shows that the docking poses of UQ establish interactions with some of the most important residues identified by the mutagenesis studies.
Based on the functional, mutational, computational, and sequence analysis carried in this project, we propose a novel UQ-binding motif, TH (XWX 2 AX 4 FX 4 )-coil-TH (XNXALX 2 -RAFXFFXFPX 2 ), that seems unique to the Na ϩ -NQR family and is not found in any other respiratory or photosynthetic enzyme. A characterization of the specific interactions of UQ with subunit D will be required to complete the understanding of this site and the residues that establish the critical interactions in the motif. Further studies are necessary to test the presence of this motif in other proteins, especially those in the genomic database that have not been assigned a function or role in the cell.
Understanding of the Catalytic Mechanism of Na ϩ -NQR-The results obtained in this report clarify several aspects of the structure and catalytic mechanism of Na ϩ -NQR. For instance, the reported crystallographic structure of Na ϩ -NQR lacks UQ, which has limited the understanding of the electron transfer chain from the structural point of view. The kinetic characterization of Na ϩ -NQR indicates that the three-and five-electronreduced states of the complex are directly involved in the catalytic mechanism (25), whereas the oxidized state of the enzyme does not play a significant role. Moreover, Na ϩ -NQR follows a Hexa-Uni Ping Pong mechanism, in which each of the substrates reacts, or is transported, independently. This mechanism is fundamentally conformational in character, and up to seven different conformations are expected (25). Under this mechanism, the substrate binding sites are not preformed and appear, or are opened, transiently in different stages of the cycle. In particular, the UQ binding site would not be opened until the five-electron reduced form binds sodium on the cytosolic side of the membrane and releases it on the periplasmic side (25). Because the crystal structure of Na ϩ -NQR was obtained in the absence of reducing agents, it is now evident that it does not correspond to any of the active forms or structural intermediates, explaining the lack of the UQ site in the structure. In addition, the distances between the cofactors seem too large to sustain physiologic electron transfer rates (12) (Fig.  5G). One of the interesting aspects of the results found in this work is that direct interactions between UQ and some of the critical residues for binding, such as Phe-213, were not identified by docking methods. Instead, other residues, such as Phe-185, which showed less importance for the activity (Table 1), appeared in direct contact with UQ (Fig. 5, C and F). The data suggest that the UQ pocket in the crystal structure is not found in its fully active conformation. Our hypothesis is that the site identified in this study is superficial and only partially opened, and in certain stages UQ might gain access to deeper parts of the protein, allowing a faster electron transfer.
The kinetic mechanism of Na ϩ -NQR also helps to explain the inhibitory behavior of HQNO and korormicin, which do not behave as pure competitive inhibitors. According to our results, these two inhibitors can indeed interact with the UQ binding site, competing for UQ and the product UQH 2 for the catalytic site in two conformations or redox states, explaining the competitive and uncompetitive components of mixed inhibition (25) and the effects of the mutations on the affinity for HQNO.
The location of the catalytic UQ site in the interface of subunits B and D is consistent with the orientation of UQ in the membrane. Studies indicate that UQ lies between the phospholipid bilayers, with the benzoquinone head fully buried in the membrane (41,42). The UQ binding pocket of Na ϩ -NQR is found in the core of the membrane (Fig. 5A), fully accessible to the hydrophobic environment where UQ is embedded, allowing a quick diffusion of the substrate and consequently a high turnover rate. In contrast, the location of the catalytic UQ site in subunit A would require the protrusion of the benzoquinone head into the hydrophilic environment, which would slow down by several orders of magnitude the turnover rate and would require energy expenditure to pull UQ out of the membrane. However, it is possible that subunit A undergoes significant conformational changes that could bring it in close contact with the lipid environment. Interestingly, the site that we are proposing is very close to the location of the covalently bound FMN in subunit B, which delivers the electrons to riboflavin. Indeed, Phe-213 points directly to FMN, within 5 Å (Fig.  5B). This distance is consistent with a high electron transfer rate observed in Na ϩ -NQR. However, the putative location of the riboflavin cofactor in the crystal lies over 35 Å away (Fig.  5G), which would not allow electron transfer from FMN in subunit B or to UQ. It should be pointed out that the superficial location of riboflavin is not supported by multiple lines of evidence that indicate that it must be deeply embedded in the protein (5,10,13), but further studies are required to confirm this hypothesis.
Conclusion-The data found in this report pinpoint for the first time the location of the catalytic UQ binding site in the structure of Na ϩ -NQR, which is a critical step to understand the structure-function relationships and the role of different residues in the catalytic mechanism of this essential bacterial respiratory complex.

Experimental Procedures
Mutant Construction-Mutagenesis reactions were performed as reported previously, using the QuikChange site-directed mutagenesis kit, with the wild-type nqr operon cloned into a pBAD expression vector as template (25). The primers designed to mutate each of the selected aromatic residues to alanine are listed in Table 5. The mutant nqr operon cloned into pBAD vector was introduced into a V. cholerae deletion strain lacking the genomic nqr operon (⌬nqr) and was used to express the mutant enzymes (25). Mutations were verified by DNA sequencing.
Activity Measurements-NADH DH and UQ RED activity were measured spectrophotometrically at 340 nm (⑀ NADH ϭ 6.22 mM Ϫ1 cm Ϫ1 ) and at 282 nm (⑀ UQ-UQH2 ϭ 11.8 mM Ϫ1 cm Ϫ1 ), as reported previously (18). NADH OX was measured as the difference between NADH DH and UQ RED . Saturation kinetics of the UQ-dependent activity were measured at different concentrations of UQ-1 (1-50 M), under nearly saturating concentrations of the two co-substrates NADH (250 M) and NaCl (50 mM). Experiments were performed in buffer containing 50 mM HEPES, 1 mM EDTA, 5% glycerol, 0.05% DDM, pH 7.5 (25). Na ϩ -NQR Reduction Assays-The reduction of wild-type and mutant Na ϩ -NQR was examined using 500 M NADH or 1 mM UQ-1H 2 as reducing agents, for the forward and reverse reactions, respectively. The purified enzyme samples were dissolved in the buffer described above to a final concentration of 50 M. Additionally, superoxide dismutase (5 units/ml) and catalase (2 units/ml) (Sigma-Aldrich) were added to the sample to produce a microaerobic environment (within 30 min) and eliminate oxygen radicals formed during Na ϩ -NQR reduction (16). Moreover, UQ-1H 2 -reduced samples contained 5 mM DTT, which guarantees full UQ-1H 2 reduction. The absorption spectra (360 -700 nm) of samples were recorded before the addition of the reducing agent and were compared with the spectrum after a 1-h incubation. UQ-1H 2 was prepared and stored as reported previously (43).
UQ-1 Binding Assays-To characterize the interaction of UQ and HQNO with the mutants and wild-type Na ϩ -NQR, binding assays were performed in samples obtained after nickel-nitrilotriacetic acid and DEAE purification with the detergent DDM. Additionally, the enzyme preparations were washed with the detergent LDAO (lauryldimethylamine oxide) to remove the tightly bound UQ-8 (27). Purified samples were diluted 1:50 in buffer containing 0.1% LDAO, incubated in ice for 30 min, and reconcentrated. The samples were diluted in DDM containing buffer (1:20) and concentrated. Two rounds of 1:20 dilutions and reconcentration steps were carried out to eliminate the excess of LDAO. Equilibrium dialysis was carried out as reported previously (30). Briefly, Na ϩ -NQR preparations (50 M) were incubated overnight in Pierce 96-well microdialysis plates (10,000 molecular weight cut-off) in the presence of different concentrations of UQ-1 or HQNO (0.1-20 M). The concentrations of the two compounds in the internal and external dialysis chambers were measured spectrophotometrically to calculate the dissociation constant (K D ) of the ligand-protein interaction. HQNO concentration was measured at 346 nm, using a molar absorptivity of 9.6 mM Ϫ1 cm Ϫ1 . The UQ-1 concentration was measured at 282 nm using a molar absorptivity of 14.9 mM Ϫ1 cm Ϫ1 (30). Under these conditions, the enzyme's activity and flavin content remained unaffected.
Protein and Ligand Models-The sequence and crystallographic structure of V. cholerae Na ϩ -NQR were obtained from the Protein Data Bank (entry 4P6V). MODELLER version 9.14 (44) was used to construct complete models of subunits B, D, and E, in particular of loops that were missing in the crystallographic structure. A template search with BLAST and PSI-BLAST (45) did not find 3D structures homologous to the missing loops. Thus, the loops were modeled based on a template from the pGenTHREADER server (46), which contains a method for fold recognition and identification of distant homologues. UQ-1 was downloaded from the ZINC database (accession number 1559692). The ZINC database (47) contains over 35 million purchasable compounds in ready-to-dock 3D formats. The atomic charges of all molecules in the ZINC database were calculated by the semiempirical quantum mechanics program AMSOL (48).
Docking-To prepare the model for docking with UCSF DOCK version 6.6, a molecular surface of the homology model with hydrogen atoms removed was prepared using DMS, a tool GGCCGTGCTTTCCTGTTCGCGGCTTACCCAGCACAGATC Y216A CTTTCCTGTTCTTTGCTGCCCCAGCACAGATCTCAG within DOCK 6.6. Another DOCK 6.6 tool (49), sphgen, was used to generate vacancy spheres surrounding the entire protein. Spheres had a minimum and maximum radius of 1.0 and 5.0 Å, respectively. UQ-1 was docked into all spheres using flexible docking. Docking orientations were ranked based on a molecular mechanics-like scoring function known as the grid score. Hydrophobic interactions between the ligand and the protein were visualized using Ligplot ϩ (50).